Fiscal Policy in Good Times and Bad Times:Distributive motivations and pro-cyclical spending∗

Martın Ardanaz† Pablo M. Pinto‡ Santiago M. Pinto§

October 20, 2013

Abstract

We explore how public spending responds to changes in the business cycle. We de-velop a political economy explanation of pro-cyclical spending where governments’incentives to deviate from a sustainable fiscal policy result from political conflicton the level of redistribution that results from a country’s position in the interna-tional division of labor. This underlying distributive conflict forces governmentswho represent more vulnerable economic actors to engage in more, rather thanless, pro-cyclical spending. We assess the plausibility of the argument using dataon government spending in Argentina over one hundred years of history. Given thestructure of the data we fit an error-correction model and estimate the short-termelasticity of government spending to the business cycle while controlling for thelong-term relationship between both series. We supplement the ECM tests withinstrumental variable estimations. The results from the ECM and IV models pro-vide support to our political economy explanation: while democratic governmentsof all shades run pro-cyclical fiscal policies, the elasticity of public spending to thebusiness cycle is greater under Peronist leaders. Alternative explanations based onideology, institutions and external conditions seem to play a lesser role.

∗Prepared for IPES, Claremont, CA, October 25-26, 2013. The information and views expressedin this article are those of the authors, and do not represent those of the Federal Reserve Bank ofRichmond, the Federal Reserve System, the Inter-American Development Bank, its Board of ExecutiveDirectors, or the countries represented by the IADB Directors.†Inter-American Development Bank; e-mail: martina@iadb.org‡Columbia University; e-mail: pp2162@columbia.edu§The Federal Reserve Bank of Richmond; e-mail: Santiago.Pinto@rich.frb.org

1 Introduction

A profuse body of literature analyzes the link between openness, income volatility and

the size of the public sector (Cameron 1978; Katzenstein 1985; Rodrik 1998). Policies

adopted under political influence as motivated by distributive concerns usually deviate

from the optimal path. Demands for government output vary across groups within the

polity, and the strength of these demands are also likely to differ along the peaks and

troughs of the fiscal cycle. In times of crises some governments are able and willing to

adopt policy responses aimed at redressing the duress created by the sharp downturns

in output, while other governments are either reluctant or unable to do so. Changes

in public spending as a result of external shocks have been a central concern in the

comparative literature explaining economic policy reform in the 1990s, and have become

a source of heated policy debates these days.1 The specific content of those responses

depend on the expected distributional consequences of the changes in relative prices

-which generate the demand for policy responses- and on contemporaneous political

conditions -which determine whose preferences are more likely to be translated into

policy.

In this paper we analyze an alternative link between redistributive conflict and eco-

nomic policy-making. We develop a novel political economy explanation of pro-cyclical

spending. We argue that political conflict arising from those distributive motivations

prevents governments from enacting fiscal policies that would smooth the movements

along the business cycle. Demands for redistribution from different socioeconomic

agents affect governments’ incentives to depart from the optimal fiscal path. We derive

those distributive motivations from actors’ position in the domestic and international

division of labor. To assess the plausibility of this argument we provide an empirical

application to the evolution of public spending in Argentina since the early 1900s.

Our explanation builds on prior research on the causes and consequences of govern-1On crises and policy reform see Cardoso & Helwege 1993; Drazen & Grilli 1993; Fernandez &

Rodrik 1991; Martinelli & Tommasi 1997; Rodrik 1996; Tornell 1995.

1

ment spending: It is fairly uncontroversial that the ability of governments to tax and

spend is affected by the business cycle. At the peak of the cycle, for instance, the fiscal

coffers are more easily filled than when the economy turns south. Economic theory

suggests that cyclical fiscal policy is sub-optimal: governments should lower spending

when the economy is growing and raise spending during hard times, if at all. While

some governments, mostly in developed countries, have followed these prescriptions

in the past, pro-cyclical behavior is far from an exceptional event as reflected in the

Southern European countries in the wake of the Global Recession of 2008, and has been

quite prevalent among developing countries. Argentina stands out within this latter

group.

The argument is built around the differential demands for government spending and

the political constraints faced by governments, which affect their ability to respond to

those demands. The incentives to engage in pro-cyclical spending –and hence deviate

from the optimal policy course– depend on the severity of the redistributive conflict in

the polity. When the incumbent government represents a coalition of actors employed

in the less productive sector of the economy, for instance, she is likely to enact tariffs and

taxes aimed at sheltering her constituents from competition and transferring resources

to them. More productive actors, usually those employed in the comparative advantage

sector of the economy, will fight against these attempts at redistributing resources away

from them. Moreover, when the coalition representing more productive actors gains

control of the government, it enacts their preferred policies of lower taxes and higher

integration into the global economy which benefit them while hurting the real income

of those actors in the comparative-disadvantage sector, leading to political mobilization

on their side. An escalation of this redistributive conflict has the potential to affect

the incumbent’s perception of her ability to retain office. As time the expectation of

staying in office falls the incumbent government spends beyond what would be optimal.2

2Moreover, shorter time horizons also affect the government’s creditworthiness in financial markets,a conjecture that we do not explore in this paper.

2

These incentives are stronger for the incumbent coalition representing workers and firms

operating in the less productive or declining sectors of the economy and would benefit

the most from the adoption of counter-cyclical policies. Our emphasis on redistributive

motivations allows us to extend and append the contributions from the extant literature

on the pro-cyclical spending which tended to focus on institutional design.

We characterize different political parties according to their links with economic

agents, which include workers and producers organized around unions, associations

and groups in different sectors of the economy. In our model the incumbent’s expec-

tation of being voted out of office is endogenous to the policies’ enacted, which in

turn depend on the government’s motivations to redistribute. Ultimately, the latter is

determined by the position in the economy of the actors that form the support base

of the incumbent party. In our model heterogeneity of preferences does matter, but

only through its effects on politicians’ expectation of staying in office. It is, thus,

equivalent to endogenizing time horizons. We explore the plausibility of the argument

using a within country time series design: We analyze the evolution of Argentina’s

public accounts from 1900 to 2004, a country whose exceptional trajectory defies tra-

ditional explanations of cyclicality. The pro-cyclical nature of government spending in

Argentina becomes stronger towards the second-half of the twentieth century, an era

characterized by stop-go cycles usually attributed to the problems that arise from the

adoption of import substitution as a developmental strategy.3

In the Argentine case deviations from the optimal policy are indeed a function of

political conflict derived from redistributive motivations that result from the country’s

position in the international division of labor as predicted by the model. Our contribu-

tion is to show that the pattern of pro-cyclical fiscal policy in the country is far from

exceptional once we account for the incentive structure faced by Argentine politicians.

We believe this is an important addition to traditional explanations based on the role3Our analysis on booms and busts of the Argentine economy follows the pioneering work of Diaz

Alejandro (1970), Dornbusch and Edwards (1991), and Rodrik (1995), among others.

3

of ideology, institutions and veto players (Velasco 2000; Kontopoulos and Perroti 1999;

Tsebelis and Chang 2004).

Collecting data on business cycles in a country like Argentina and documenting the

nature and depth of the shocks allows us to conduct empirical analyses on whether

the pro-cyclical nature of Argentina’s fiscal policy is indeed a function of political

conditions. Given the structure of the data, we fit an error-correction model which

allows us to estimate the short-term elasticity of government spending to the business

cycle, controlling for the long-term relationship between both series which determines

the rate of adjustment for short term deviations from the trend that binds the two series.

Implementing these models is fraught with complications. Among them, the ability of

the government to finance its spending, through taxation, loans or monetization of the

deficit, depends on the performance of the economy. Hence it is important to estimate

the effects of economic shocks on aggregate spending, or even on sectoral spending

to the extent possible. We also need to account for political economy motivations of

spending: while in times of crises most groups would prefer being bailed out, in normal

times actors hold different preferences with respect to the optimal level of government

intervention in the economy. In both instances, however, economic agents are far from

indifferent on who bears the burden more heavily. This justifies the focus on a within

country analysis and the emphasis on the cyclical nature of spending, rather that its

levels.

Identifying off of the short and long-run relationship between spending and the

business cycle allows us to assess how the nature of the ruling coalition is more im-

portant than the ideological stance of the incumbents in explaining changes in public

spending, the dependent variable in our tests.4

4We establish the occurrence of thirteen economic shocks affecting the country from the early 1900sthrough the Global Financial Crisis of 2008. These instances of crisis include short medium termrecessionary periods of the business cycle, lasting as long as 4 years. They are: 1913-1917; 1929-1932;1951-1952; 1958-1959; 1961-1963; 1974-1976; 1977-1978; 1980-1982; 1984-1985; 1987-1990; 1994-1995;1998-2002; and global recession of 2008. Data on the crises was obtained from CEPAL and FIEL.See Table B in the Appendix. While some of these shocks to economic output resulted from drops indemand, changing terms of trade, financial recessions and openness at the global level, elsewhere we

4

We present results from statistical analyses contrasting our argument of coalitional

determinants of pro-cyclical spending with alternative explanations that emphasize the

role of institutions, ideology and discourse, and external motivations and constraints.

We find evidence suggesting that the political economy hypothesis of cyclical fiscal be-

havior is plausible: while governments of different shades run pro-cyclical fiscal policies

at times, the elasticity of public spending to the business cycle is greater under the

Peronist. The incentive to redistribute away from the comparative advantage sector is

what sets the Peronist apart from other political coalitions. The catch-all nature of the

party -binding together a slew of urban workers, industrialists and traditional sectors in

the interior of the country- not the dynamics of coalition bargaining in the legislature

is what shortens the leaders’ time-horizons and reduces the government’s ability to tap

financial markets, thus resulting in more pro-cyclical spending. Our results are robust

to instrumental variables estimation where the output gap of Argentina is instrumented

with the output gap of the six other major economies in the region (Ilzetzki and Vegh

2008; Jaimovich and Panizza 2007; Gali and Perotti 2003). Alternative explanations

based on ideology, openness and institutions seem to play a lesser role.

2 The politics of booms and busts

How do governments choose their fiscal policy in response to aggregate economic

shocks? Governments can rely on different policy instruments to deal with output

fluctuations along the business cycle. Fiscal policy is one of the tools that governments

can deploy to smooth externalities generated by “bad” states of the economy. In the

classic work of Musgrave (1959), stabilization is one of the three objectives or functions

of fiscal policy in modern economic systems. 5 By studying public spending decisions

during different phases of the business cycle, scholars assess the extent to which fiscal

show that unsustainable fiscal policies, macroeconomic instability, currency and banking crisis at thelocal level played a major role (Pinto 2013).

5The other two being efficiency and redistribution.

5

policy performs this task.

According to several important strands of economic theory, fiscal policy should be

counter-cyclical (or at least acyclical) to perform its stabilization function: spend-

ing should be increased during recessions in order to stimulate aggregate demand and

protect vulnerable groups, especially the poor.6 During expansions, the government

should reduce spending in order to “cool off” the economy and contain inflationary pres-

sures. While this “benevolent planner” prescription is followed to a great extent among

OECD countries, fiscal policy is often pro-cyclical (Kaminsky et al. 2004; Alesina et al.

2008). That is, total government spending as a share of GDP goes up during booms

and down in recessions, a policy pattern that adds to macroeconomic instability and

hurts the poor the most since they lack the assets or access to financial markets that

would allow them to smooth out adverse income shocks. Previous studies have shown

that pro-cyclicality is a particularly acute problem in developing countries, especially

those in Latin America (IDB 1997; Gavin and Perotti 1997). Using a regression based

measure of fiscal cyclicality from Woo (2009), Figure 1 shows that Latin America has

the highest levels of pro-cyclicality around the world. Yet this figure does not capture

the substantial variation in cyclicality within each group of countries that has been

documented in the literature. Given these stylized facts, the challenge is identifying

which factors account for variation in the business cycle properties of public spending.

Or put differently, why do countries differ in the way fiscal policy reacts to movements

along the economic cycle, or even aggregate economic shocks?

The classical explanation of pro-cyclical behavior is grounded in the existence of

credit constraints. In a seminal contribution to this literature Gavin and Perotti’s

(1997) argue that developing countries find it hard to follow countercyclical policy

because they lack access to international credit during recessions. The problem with

this explanation is its inability to provide answers to the following: why are countries6Both Keynesian macroeconomic policy and tax smoothing arguments (Barro 1979) prescribe coun-

tercyclical fiscal policy.

6

Figure 1: Government spending and GDP

not able to self-insure by accumulating reserves in good times? Moreover, what keeps

lenders from providing funds to countries when access to borrowing helps them smooth

out the cycle?

A similar problem is found in arguments linking pro-cyclical policy to the nature

of the tax base in developing countries (Talvi and Vegh 2005) or patterns of integra-

tion with the world economy (Wibbels 2006). Talvi and Vegh first note that output

(consumption) in developing countries is twice (three times) as volatile as in the in-

dustrial economies. Given that developing countries often base their tax systems on

consumption taxes, their tax bases are considerably more volatile than those in the

OECD. Developing countries find it harder to follow countercyclical fiscal policies since

the problem associated with the volatility of their revenue streams is compounded by

the many political pressures to spend in good times. Similarly, Wibbels (2006) argues

that increasing trade exposure has exacerbated the volatility that complicates counter-

7

cyclical social spending in the developing world. Yet these explanations are far from

satisfactory: in theory, output volatility and trade openness are neither necessary nor

sufficient conditions for pro-cyclical spending. If anything, volatility and trade openness

should point in the opposite direction: the more volatile and exposed to international

trade an economy is, the higher the incentives for politicians to behave in a counter-

cyclical way, by providing social insurance, creating instruments such as stabilization

funds, or allocating higher shares of automatic stabilizers in the budget. After all, that

is the lesson learned from the experience of small and open economies in the developed

world (Cameron 1978; Katzenstein 1985; Rodrik 1998).

Given the limitations of purely economic explanations, scholars have turned to the

study of the political economy determinants of pro-cyclicality. As Talvi and Vegh elo-

quently put it: “The root of unstable policies may lie, not in policymakers inability to

set the “correct” policies, but rather in the political economy of fiscal arrangements.”

(Talvi and Vegh 2005, p.159). A political economy perspective can provide insights to

explaining why some countries and governments are more pro-cyclical than others, or

even why there is over time variation in the extent of pro-cyclicality within a country,

as is our interest in this paper. While several competing arguments have been pro-

posed within this perspective, they all share a common focus: the study of the factors

pertaining to the political arena (e.g. political institutions, elections, coalitions) that

affect economic policymaking and fiscal outcomes in systematic ways (Drazen 2000).7

The rest of the section briefly surveys this emerging literature.

2.1 Political economy of pro-cyclical spending

Political economy explanations of fiscal outcomes build on the idea that fiscal decisions

are the result of political processes that involve actors with varied interests. These7The bulk of the political economy literature on the determinants of economic policy outcomes

usually concentrates on the levels –such as the effect on electoral rules on the size of government–rather than the cyclical properties of public spending, which is the focus of our paper. Thus, we donot review this important strand of the literature here but refer to Persson and Tabellini (2003) andEslava 2006 for extensive surveys.

8

interactions take place mainly between politicians and voters, or among politicians who

represent diverse interests or constituencies. In this tradition, scholars have identified

a number of political distortions that tend to generate a pro-cyclical bias in fiscal

policy. These distortions can be grouped in two types of problems: “cooperation” and

“agency” problems.

Cooperation problems. A classic example of failures in cooperation is the well-known

common pool problem (Ostrom, 1990).

In fiscal policy, the common pool is the budget that political players draw upon

(financed from a general tax fund) to generate concentrated benefits (such as targeted

public policies). Tornell and Lane (1999) develop a model in which multiple political

groups compete for a share of the common pool, leading to a “voracity effect”: a more

than proportional increase in spending in response to shocks, such as a terms of trade

windfall. Similarly, Talvi and Vegh (2005) present a model in which abundant fiscal

resources create pressures to increase public spending. In these models, the voracity

effect is simply assumed, but not analytically derived. They provide little guidance to

understanding the magnitude of the problem. What factors determine the intensity of

the voracity effect, and hence, the level of pro-cyclicality?

First, the number of actors drawing from the common pool could be construed as a

determinant of the voracity effect. The pressure to overspend during upturns increases

as the number of groups drawing from the common pool increases. Braun (2001) and

Lane (2003) find evidence consistent with this hypothesis for developing and OECD

countries respectively: in both papers, as the number of political veto players increases,

fiscal policy becomes more pro-cyclical.

In addition to fragmentation, political polarization has also been hypothesized as

a key determinant of pro-cyclicality (Humphreys and Sandbu 2007; Ilzetzki 2009; Woo

2009). The intuition is that as the preferences over the desired distribution of public

spending between political groups diverge (or more generally, the deeper the division

9

prevalent among the groups), the greater will be the incentive of policymakers to spend

too much while in power, leading to pro-cyclical fiscal policies.

Principal-agent problems. In models based on problems of cooperation reviewed

above, the role of political demand by voters and groups is theoretically underspecified.

This omission is particularly problematic if one is concerned about the endogenous

nature of political outcomes: expenditure decisions can affect the politician’s likelihood

of remaining in power. If such consequences could be anticipated, politician’s are

expected to modify their behavior accordingly. Moreover, these models seem to neglect

a basic tenet of electoral competition: that politicians devote public resources to remain

in office and incumbents may engage in fiscal manipulation (e.g. political budget cycles)

to influence voters and retain power, a practice that is pervasive among developing

countries (Ames 1987; Brender and Drazen 2004; Schuknecht 2000; Shi and Svensson

2006).

To explicitly account for the endogeneity of electoral outcomes, Alesina et al. (2008)

develop a political agency model that brings voters back into the picture. The model

is used to interpret their main empirical finding: a positive correlation between pro-

cyclical policy and measures of corruption. Alesina and co-authors argue that pro-

cyclicality is the response by rational voters who operate in an environment charac-

terized by information asymmetries when facing corrupt governments.8 Under these

conditions voters demand higher transfers to themselves (in the form of public goods)

during good times to “starve the Leviathan” when economic activity drops. Faced with

these pro-cyclical demands, governments do not accumulate reserves during booms; on

the contrary, they incur in large debts which prevents them from being able to finance

spending in bad times. What remains unclear from this model is why voters, given that

they don’t trust the government, are more (rather than less) likely to demand public

goods and accept a higher level of rent extraction by politicians.

Our argument builds on the insights from the “cooperation” tradition. Departures8Voters observe the state of the economy, but cannot observe government borrowing.

10

from the optimal fiscal path are endogenous to political conflict that results from dif-

ferent redistributive motivations among different agents. We expect the voracity effect

should be sharper under democratic competition because the incumbent needs to com-

pensate a high number of potential veto players. Demands for government spending

are higher when the government relies on the support of a coalition base of workers and

businesses owners operating in less productive sectors of the economy (the protectionist

coalition). The incentives to spend in good times are exacerbated when the probabil-

ity of retaining office depend on the policy choices made by the ruling coalition. The

corollary is that governments representing economic actors who would benefit the most

from countercyclical spending are paradoxically more likely to behave pro-cyclically.

3 A political economy explanation of pro-cyclical fiscal

policy

In this section we discuss the intuition from a political economy explanation of cyclical

spending. We combine issues traditionally associated with the trade theoretic litera-

ture, such as the relationship between productivity differentials and the expansion or

contraction of productive sectors, with elements of the rich literature on the political

determinants of fiscal policy discussed in the previous section.9

In the model there are four actors: government, two types of firms, and workers

interacting over two periods. Firms (A) and (C) receive an endowment of capital in pe-

riod t=0, which they may choose to consume or invest in two different activities/sectors

of the economy: a riskless activity X, or a risky activity M .10 In the first period the

government may choose to tax the endowment of capital available in each sector, and

distribute that revenue among any actor in the polity. The amount available for invest-9See Alesina and Perotti 1995 and Eslava 2006 for reviews.

10This would be equivalent to assuming that the country has a comparative advantage in sec-tor X (agricultural sector), or more broadly sector X is relatively more competitive than M (ur-ban/manufacturing sector). We label the sectors to match the traditional pattern of comparativeadvantage observed in Argentina.

11

ment is the difference between the original endowment, minus consumption and taxes.

We divide firms into two groups: a group of more productive firms who could operate

in both sectors M and X and less productive firms who can only enter the riskier sector

M . Firms in activity X do not require government support to be competitive, while in

activity M only a fraction of more productive firms survive, while others would require

government transfers, subsidies, and/or trade restrictions to stay in the market. La-

bor is primarily employed in activity M ; wages and employment levels are negotiated

between the firms in sector M and a labor union prior to the realization of a random

variable µ which determines the minimum level of productivity required to to remain

active in the sector.11 Firms below that threshold are liquidated. If the firm is liq-

uidated, then labor hired by the firm becomes unemployed, while investors can still

recover part of the value of capital invested.

After being elected as a result of a probabilistic voting process, the government

can choose among the following instruments: taxes and transfers which could vary

across sectors, a transfer to labor, and unemployment benefits. The government faces

a budget constraint: transfers and benefits must be less than or equal to taxes collected

on firms’ income and wages.12

Governments are partisan in the sense that they weigh more heavily the well-being of

some actors in the polity. Partisan governments, thus, face different political pressures

for additional spending. Specifically, there are two types of governments Gx, where

x = {R,D}, representing socioeconomic groups in the more productive and the less

productive sectors respectively. Workers are represented by a government of type GD,

who prefer a higher provision of transfers to labor, and capitalists are represented by

a government of type GR, who prefer a lower level of transfers to labor.11We make the simplifying assumption that labor is employed in sector M . Alternatively we could

have assumed that labor is employed in sector X or that there are two types of labor: skilled workers,who are more intensively employed in sector X, and unskilled workers, who are more intensivelyemployed in M .

12In an extension to the model the government may access to credit markets, which allows for publicinvestment or consumption above and beyond what is collected from taxes.

12

The government in power in the first period collects taxes and determines the level

of transfers targeted to the political groups. She will also choose the corresponding

taxes and transfers in the second period if she remains in power power. Otherwise,

the incumbent is replaced by a rival party who decides on her preferred level of taxes

and spending/transfers. The incumbent’s probability of staying in office is endogenous

and determined through probabilistic voting. The government, thus, faces a tradeoff

between catering to her constituents, increasing her indirect utility, or catering to other

groups in the polity which increases the probability of staying in office.

The level of current expenditures tends to increase as the expectation of being re-

moved from office increases (or the probability of staying in office drops). A simple

comparative statics exercise with respect to government type shows that the correlation

between output shocks (through changes in the endowment or changes in productivity

of the endowment) and government spending is not uniform for all types of govern-

ments: it depends on the government’s motivation to redistribute, which is a function

of its constituency base. The elasticity of spending to economic shocks is above unity

positive for the type-GD, but not for governments of type-GR. To the extent that a

type-GL government requires increasingly larger transfers from other sectors to satisfy

the demands of its constituencies, political tensions with actors in the comparative

advantage sector eventually arise. In sum, incentives to extract from productive firms

reduces the probability that a government of type D will remain in office. the threat of

losing office makes government representing the actors who would benefit from counter-

cyclical spending less willing to defer expenditures and reduce the demands on the

productive sectors of the economy. Moreover, as the type D government’s demands for

redistribution increase, the reaction from the opposing camp becomes stronger, further

lowering the expectation that the incumbent will remain in office.

13

3.1 Redistribution and credit constraints

The relationship between economic shocks and spending patterns may be exacerbated

by credit market imperfections in ways that are not captured by the model. Such imper-

fections reinforce the pro-cyclicality generated by political distortions. With complete

markets, the optimal fiscal policy consists in completely smoothing out government

consumption. The demand for consumption smoothing will be particularly stronger

for actors who have no ability to access financial markets, such as workers and less

productive firms –i.e., those with a higher probability of failing. However, policymak-

ers have typically faced a loss of confidence and thus intensified borrowing constraints

during bad macroeconomic times; this problem is particularly pronounced in devel-

oping countries. The inability of accessing financial markets further constrains the

government’s ability to run a counter-cyclical fiscal policy (Gavin and Perotti 1997).

The problem is exacerbated under conditions of high levels of political conflict and

shortened time horizons.

The fiscal policies implemented by different types of governments may lead to dif-

ferent degrees of public creditworthiness. Specifically, consider a capital market imper-

fection that generates a positive association between public debt and the risk premium

(Aizenman, Gavin, Hausmann, 2000). If the risk premium is higher in bad times,

then borrowing in bad times becomes more costly. Policymakers, as a result, would

be induced to raise taxes and reduce government consumption, further fueling the pro-

cyclicality of the fiscal policy. In good times, the opposite would take place: a lower

risk premium would encourage more borrowing, more government spending, and lower

taxes.

3.2 Testable Hypothesis

Combining these propositions we can derive the following testable hypothesis:

14

The correlation between spending and output will be greater when the incumbent

represents workers and firms in the comparative disadvantage sector of the econ-

omy.

In the Argentine case we would expect the coalition of workers and industrialists in

the urban areas built around the Peronist Party are more likely to run pro-cyclical fiscal

policies, leading to a higher correlation between spending and output. There are sev-

eral alternative reasons of why coalitions representing workers and industrialists, such

as the Peronist party in Argentina, may face more difficulties in keeping government

expenditures under control. First, these types of governments are subject to more fiscal

pressure; second, members of the coalition disagree on the level of wages and labor ben-

efits. In general, cooperation is more difficult to achieve when the number of players is

large (Bawm and Rosenbluth 2006); yet the problem is compounded when the prefer-

ences of those players are not fully aligned. Second, individual coalition members have

veto power, and can block a proposal of another coalition member. At the same time,

each subgroup has little power to implement its own program by going alone. Finally,

enforcement mechanisms among coalition members will be relatively weak when the

members of the coalition expect a higher probability that the government representing

them will be removed from office. A short tenure will limit the possibilities to play the

repeated decision-making game among the political players. The members’ incentives

to cooperate are therefore further reduced.13 We need not assume that coalition gov-

ernments necessarily have a shorter tenure than have one-party governments; this result

is derived from the argument. Distributional motivations faced by the incumbent gov-

ernment, not the number of coalition members, drive the result. While workers prefer

counter-cyclical policies, governments representing them place more demands on the

competitive sectors of the economy, hence increasing the potential for conflict, which

in turn makes spending more pro-cyclical.13On the role of time horizons in fostering (or hindering) political cooperation, see Alesina (1988)

and Dixit et al (2000). Spiller and Tommasi (2007) provide an application of this framework in theArgentine institutional context.

15

Figure 2: GDP per capita and pro-cyclical government spending

4 Empirical strategy: Analysis of fiscal policy in Argentina

As discussed in section 2 pro-cyclicality is an acute problem, especially among de-

veloping countries. Figure 2 presents the well-known correlation between economic

development and fiscal outcomes in a sample of 100 countries.14 An interesting point

that emerges from figure 2 is that Argentina is: first, much more pro-cyclical than

would be expected for its level of economic development; and secondly, that is well

above the average level of pro-cyclicality.

Figure 3, which is based on a standard methodology to calculate fiscal pro-cyclicality,

presents a long run view of the cyclical behavior of public spending.15 In countries like

Argentina we observe the co-movement of both lines: fiscal behavior seems to follow

the economic cycle. Yet this figure ignores the short-run variance around the trend,14The data on pro-cyclicality used in figure 2 is from Kaminsky et al (2004).15This measure is the correlation of the cyclical components of government spending and output that

are first filtered by the Hodrick-Prescott (HP) method.

16

Figure 3: Government Spending and Economic Output in Argentina

which for the Argentine case is reproduced in Figure 3. It can be seen that the cyclical

components of output and spending tightly covary; we can also see that in most years

the proportional change in spending is greater than the proportional change in output.

Figure 4 graphs the deviations around the trend in US output and public spend-

ing, reflecting the countercyclical spending pattern that is common among developed

countries (Lane 2003). The US behavior stands out in stark contrast to the pattern

of pro-cyclical spending in Argentina presented in Figure 3. The equivalent graph on

the pro-cyclical behavior of social spending in Argentina since 1980 is reproduced in

Figure 5. The graph not only suggests that social spending has been cyclical, but also

that changes in spending seem to be magnified at each stage of the business cycle.

Argentina stands out as one of the countries with most pro-cyclical behavior in

the world. Argentina’s pro-cyclical fiscal behavior has been documented in numerous

studies including Sturzenegger and Moya (2003) and is traditionally attributed to the

17

Figure 4: Social Spending and Economic Output in the US

adoption of unsustainable policies as a consequence of an economic strategy of selective

protectionism aimed at development of the manufacturing sectors; the outcome is an

ever increasing growth of the government sector (see Figure 6).16

The strategy would allegedly transcend institutional settings and partisan alle-

giances pushing the economy on a downward spiral path. Prior research suggests that

the first failed efforts to reform and liberalize the economy can be tracked backed to the

military regime of 1976; economic reform ultimately succeeded under the leadership of

Carlos Menem and the Peronist party in the 1990s (Sturzenegger and Moya 2007). Yet

figure 7 and Table 1 show that federal government spending became more rather than

less pro-cyclical in the 1990s. Moreover, Figure 6 also shows that government spending

and revenue seem to covary more tightly together since the mid-1970s. We can see in

Figures 3 and 5 that output and spending seem to covary in the short-term even in the16The higher ratio of the public sector to output is also consistent with Wagner’s Law of increasing

state activity. There is a vast body of empirical work that unveils this relationship.

18

Figure 5: Social Spending and Economic Output in Argentina (1970-2004)

Figure 6: Government Spending and Revenue

19

Figure 7: GDP and Spending: Correlation by Decade

1990s, the era of structural adjustment and reform. Cyclical spending in the Argentine

case does not seem to be a legacy of ISI.

In Table A in the Appendix we schematically present the main characteristics of

each Presidency in Argentina since 1900. Our expectation is that the Peronists, who

have traditionally represented urban workers and industrialists in the less competitive

or comparative disadvantage sector of the economy, are more likely to spend in the

upswing of the cycle.17 The Peronists’ redistributive motivations that result from

their strong attachment with their political base and the rift with the coalition of

agents in the comparative advantage/agricultural sector makes Peronist governments

less creditworthy in the eyes for lenders and hence they find a tougher time financing17 The Peronist is a party representing an urban coalition of workers and businessmen in manufac-

turing and traditional sectors in the Center, North and Northwest of the country (on the support baseof Peronism, see Smith 1972; Mora y Araujo and Llorente 1980; Calvo and Murillo 2004). The mili-tary, on the other hand, who ruled the country intermittently since the 1930s have been the politicalsuccessors of the Conservatives. The so-called Military Party tended to represent the interests of landowners in the Pampas and the temperate agricultural producers of grain and beef producers.

20

government spending in the down cycle. Thus, we expect the Peronist to engage in

more pro-cyclical spending than governments of different orientation and constituency

bases. 18 In the ensuing sections we will test our hypothesis using time series data from

Argentina. Our empirical strategy below aims at explaining the correlation between the

covariance of the cyclical components of both series as a function of the characteristics

of the ruling party. We move now to assess the empirical content of our main hypothesis.

4.1 Statistical models and results

As discussed in earlier sections Argentina stands out as being a country with a history of

pro-cyclical fiscal policy. We are interested explaining the variance of the pro-cyclical

behavior of spending at the federal level over time. In particular we would like to

estimate whether and how political conditions -namely the nature, orientation and

characteristics of the governing party- affect the cyclicality of government spending to

economic output. A positive correlation would suggest that spending is pro-cyclical, a

negative relationship would suggest that spending is anti-cyclical, while no statistical

relationship between the series would be evidence of acyclical behavior.19 Note that

even after the log transformation value added and government spending, like many

economic variables, are not stationary.20 Both series seem to be integrated of the18The other main traditional party in Argentina is the UCR. The Radicals are usually associated

with the middle class. From Yrigoyen to Alfonsin the Radicals were in permanent confrontation withagricultural interests (de Alvear and de la Rua less so), and were never never able to gain the fulltrust of workers, unions and industrialists. Despite its confrontational attitude towards agriculturalproducers, we do not expect the Radicals to face the same incentives as the Peronist. There were twoinstances when Radical leaders attempted to engage the base traditionally represented by the Peronist;in both instances they ultimately failed. Arturo Frondizi’s ascent to power in 1958 was the result ofa secret pact with Peron, who was in the exile at the time; his party was proscribed by the militarythat ousted Peron and could not field candidates in the election. Frondizi’s developmental policieswere a clear attempt at gaining the hearts of workers; yet his relationship with unions remained hostilethroughout. After failing in his attempts to weaken the Peronist stronghold of trade unions, in 1984Raul Alfonsin appointed a unionist, Hugo Barrionuevo, to the Minsitry of Labor and Social Securitywith mixed results: the overture gave Alfonsin the opportunity to successfully launch the Plan Australin 1985. Yet the union leaders in his cabinet were never able at curbing organized labor’s belligerence.

19For ease of interpretation we take the natural log of both variables (see below for a derivation ofthe estimating equation).

20We chose to normalize spending and GDP by population. Using the series in their original dollarvalue yields identical results.

21

first degree -i.e. I(1)- and hence stationary after first-differencing.21 Fitting an Error

Correction Model allows us to estimate a co-integration parameter that accounts for the

long-term/equilibrium relationship between both series. We are also able to estimate an

adjustment parameter that determines how fast spending needs to adjust when either

variable moves out of equilibrium. The error correction model assumes that spending

and output are in a steady state equilibrium which can be represented as:

∆Gt = µ + β ∆Yt + γ (Gt−1 − δ Yt−1) + εt (1)

The second term in (1), β ∆Yt, can be interpreted as the short term effect of output (Yt)

on government spending (Gt). The third term, γ (Gt−1 − δ Yt−1), captures the long-run

relationship between spending and output. If the series are cointegrated the absolute

value of the γ parameter will be negative and statistically different from zero. This γ

parameter captures the error correction term: if for whatever reason the series are not

in equilibrium because either spending or output deviate from their trends, γ is the rate

at which spending will adjust each period to bring the series back to equilibrium. The

coefficient δ is the cointegration parameter, i.e., the long-run elasticity of government

spending with respect to output.22 If the series are cointegrated of the same order, as

they seem to be, δ can be estimated by regressing Gt on Yt using OLS; this estimate

of δ is superconsistent. The residuals of this first stage regression enter as the second

term in equation (1).23 The first model in Table 2 reproduces the results of the second

stage regression. These results corroborate that government spending in Argentina

is indeed pro-cyclical: at .82 the coefficient on change in output (the elasticity of21The augmented Dickey Fuller test statistic for the logarithmic transformation of the government

spending series is: Z(t)= -1.327; we cannot reject the null of the series having a unit root (the McKinnonapproximate p-value is .6169. First-differencing the series yields a test statistic of: Z(t)= -10.705, whichsuggests that the series in differences is stationary (with an approximate p<.0001). The natural log ofthe GDP/capita series is also I(1), with a Z(t)= -1.316 (approx. p=.6219); the series in first-differencesreturns a Z(t)=-9.123 (approximate p<.0001).

22See Engle and Granger 1987; Beck 1991; Akitoby et al. 2002 discuss an application to publicspending.

23As customary in the empirical literature we estimate the first-stage regression with a constantterm; hence the Residualst−1 = Gt−1 − θ − δYt−1. Using a one-stage model yields identical results.

22

government spending to GDP) is substantively large, and statistically significant. The

coefficient on the residuals from the first stage suggest that the error correction process

is relatively fast, with 43% of the adjustment occurring each period.24 In order to

test our hypothesis of differential partisan motivations and constraints on government

behavior leading to pro-cyclical government spending we augment equation (1) to allow

for the correlation to vary:

∆Gt = µ0 + µλ λ+ λ βλ ∆Yt + β1 ∆Yt + γ (Gt−1 − δ Yt−1) + εt (2)

The parameter λ in equation (2) is an indicator variable that takes the value of 1 if

the incumbent government is Peronist.25 Our expectation is that βλ > β1. Our ex-

pectations seem to be borne out in the results reproduced in models (2)-(4) on Table

2: there is a strong evidence that Peronist governments run more pro-cyclical fiscal

policies. Ideology, proxied for by a dummy variable for the periods under which the

prevailing developmental strategy was import substitution (1930-1976), does not seem

to play a role on the pro-cyclicality of spending.26 Spending seems to be less pro-

cyclical or even a-cyclical under Military rule which is consistent with our expectations

given their constituency base: agricultural producers in the competitive/comparative

advantage sector.27 Table (3) reproduces results from model (2) in Table 2, now ex-

cluding the Peronist administrations and adding interactions between change in output

and each of the non-Peronist administration. These results show that the short-term

elasticity of spending to output is at its lowest levels under Military rulers.28

24In additional models we find that the short-term elasticity and the error correction parameter be-come larger in the second half of the twentieth century, which is consistent with the evidence presentedearlier. We also find a strong and significant correlation between spending and revenue, esp. in the1970s. All these results are available from the authors upon request.

25Table A in the Appendix summarizes the tenure, party affiliation and ideological orientation ofArgentine Presidents since 1898.

26We tried with different start and end-dates to capture the ISI-era with similar results.27Time horizons fared prominently in the motto of the Junta in charge the autocratic regime of

1976-1983, the so-called Proceso de Reorganizacion Nacional (Process of National Reorganization):“the [Proceso] does not have deadlines, but objectives.”

28 The exception are the members of the Grupo de Generales Unidos (GOU) of which colonel Peronwas a prominent figure. This group engineered the coup of 1943 that led to the ascent of then Peron

23

In Tables (4)-(6) we present alternative specifications of the ECM models. In Ta-

ble (4) we add dummies for all democratic presidents and for the regime that paved

the ascension of Peron to power (the Group of United Officers, or GOU in its Span-

ish acronym), which had a similar support base. These results underscore the stark

differences between the leaders representing the urban coalition and those of differ-

ent orientation, autocratic or democratic alike. While under the Peronist and GOU

tenures the short-term elasticity of spending to output is substantively (above unity)

and statistically significant, the behavior of public spending under autocratic leaders

and non-Peronist democratic leaders appears to be a-cyclical. The results are robust

to the inclusion of controls for changes in the terms of trade, import quantities, the US

short-term interest rates, developmental strategies, and even time (decade) dummies.

Among the controls, it is apparent that independent of the short term changes in

output, changes in government spending were higher in the ISI and Reform eras (see

Table 5). Short-term interest rates in the US have a negative and significant effect

on spending, which suggests that access to credit markets may affect the ability of

governments to finance public spending. Changes in the terms of trade and trade

openness, on the other hand, do not seem to be related to government spending. It

is also worth noting that Table 6 returns a positive albeit small pro-cyclical behavior

under Conservative leaders; this is unsurprising given that in the absence of direct taxes

the main source of revenue at the time where specific tariffs on imports. There is no

reliable data for this period, but accounts in the secondary literature suggest that most

public spending at the time took the form of investment, not consumption, particularly

in the form of infrastructure.

In sum, we find evidence that government spending is pro-cyclical under the Per-

onist, a party that represents organized labor and import competing manufacturing

interests in urban sectors of the economy. While not a definitive test of our argument

to power. A short period of time in General J.C. Ongania’s tenure also seems to depart from thetraditional characterization of the military as pro-agriculture. His administration courted the supportof the neo-Peronist faction led by the steelworkers’ union Augusto Vandor.

24

these findings suggest that our political economy explanation is plausible.

4.1.1 Robustness check: instrumental variable estimation

So far, we have assumed implicitly that the causality goes from the business cycle to

fiscal policy. However, endogeneity issues (e.g. output reacts to fiscal policy) could in

principle be driving our results. Thus, in this section we turn to instrumental variables

as a way of dealing with this problem. Following recent contributions on the subject

(Ilzetzki and Vegh 2008; Jaimovich and Panizza 2007; Gali and Perotti 2003), we

instrument the output gap of Argentina with the output gap of the six other major

Latin American and Caribbean economies during the period of analysis: Brazil, Chile,

Colombia, Peru, Mexico, and Venezuela (LAC6).29 As shown by Figure 8, there is a

strong relationship between external and domestic income shocks.30

Table 7 reproduces results from estimating the following equation by instrumental

variables (2SLS):

Gct = α1 + α2 Yct + α3(Y c

t × λt) + α4 λt + α5 Xt + εt (3)

where Gct is the cyclical component of real government spending at the federal level;31

Y ct is the output gap (deviation of observed GDP from potential GDP in percentage

terms as defined in footnote 30); λt is an indicator variable that takes the value of

one when the incumbent is Peronist, and Xt is a vector of other controls, such as the

percentage change in the terms of trade (∆ToT ). Our main result continues to hold:

Peronist incumbents increase the degree of fiscal procyclicality.29Along with Argentina, these economies represent around 90 percent of the total GDP of the Latin

American and Caribbean region.30 The definition of output gap is standard: the percentage deviation of actual from potential GDP,

using the HP filter technique (setting the smoothing parameter, λ, at 100).31We use federal government spending rather than total government spending because the former

reflects the level at which we measure the right hand-side variables, particularly political orientationof the incumbent coalition around which our argument is built.

25

Figure 8: Output Gap Correlation: Argentina and Latin American Six

5 Conclusion

In this paper we argue that fiscal responses to responses to the upward and down-

ward swing of the economic cycle depend on the nature and orientation of the ruling

coalitions. While some of the responses are aimed at redressing the duress created

by the sharp downturns in output, incumbents’ ability to respond to economic crises

is also likely to reflect the underlying distributive cleavages in the polity. Our argu-

ment on the incentives to engage on pro-cyclical spending emphasizes the role of time

horizons which, in turn, are affected by the nature of the underlying coalition and its

incentives to redistribute. Endogenizing fiscal policy to politics allows us to extend

and append the contributions from the extant theoretical and empirical literature on

pro-cyclical spending. Using time-series data on the evolution of government spending

in Argentina, a country classified as highly pro-cyclical, we find preliminary support to

our main hypothesis.

26

The argument and findings have strong implications for our understanding how

the incentive structure faced by governments has the potential to affect governments’

ability to follow an optimal policy path. Moreover, we find that this incentive structure

dominates even when that beneficiaries from enacting the optimal policy are the core

constituents of the incumbent. While counter-cyclical spending would have benefitted

workers and other economic agents who do not have access to financial markets to

smoothen their consumption, when in power the Peronist party has engaged in pro-

cyclical behavior. Consistent with our argument the Peronist, who represent a coalition

of industrialists and workers in the comparative disadvantage sector of the economy,

have systematically used spending as a political tool when the economy expanded and

where forced to retrench spending at times of economic crises.

Last, our model brings to the fore a traditional puzzle in the literature on the po-

litical economy of redistribution: we can identify a virtuous equilibrium where winners

and losers from economic integration are able to credibly commit to a compensatory

scheme that reduces the incentives to fight and extends the shadow of the future. Under

these conditions the constraints faced by incumbents in enacting optimal fiscal policies

are reduced. The political conditions under which these compromises attain are usu-

ally attributed to differences in institutions, broadly defined (Katzenstein 1985, 1987;

North 1990). Yet institutions are endogenous to the political process. Why forward

looking actors are unable to establish those institutions that make these compromises

more likely to attain remains a puzzle that we cannot explore with the data at hand.

27

References

Aizenman, J., M. Gavin, and R. Hausmann (2000). “Optimal tax and debt policywith endogenously imperfect creditworthiness.” Journal of International Trade andEconomic Development, Vol. 9: 367-395.

Akitoby B., B. Clements, S. Gupta, and G. Inchauste (2002). “The Cyclical and Long-Term Behavior of Government Expenditures in Developing Countries.” IMF Workingpaper WP/04/02.

Alesina, Alaberto (1988) “Credibility and Policy Convergence in a Two-party Systemwith Rational Voters.” American Economic Review, Vol 78 (September 1988):796-806.

Alesina, Alberto and Roberto Perotti (1995). “The political economy of budget deficits.”IMF Staff Working Paper.

Alesina, A., F. Campante, and G. Tabellini (2008) “Why Is Fiscal Policy Often pro-cyclical?” Journal of the European Economic Association, Vol. 6 No. 5, 1006-1036.

Ames, Barry (1987). Political Survival: Politicians and Public Policy in Latin America.Berkeley, CA: University of California Press.

Barro, Robert (1979) “On the Determination of the Public Debt” Journal of PoliticalEconomy 87 (5); 940-71. Part 1, October.

Bawm, Kathleen and Frances McCall Rosenbluth (2006). “Coalition Parties vs. Coali-tions of Parties: How Electoral Agency Shapes the Political Logic of Costs and Bene-fits.” American Journal of Political Science, Vol. 50, No. 2: 251–266.

Beck, Nathaniel (1991). “Comparing Dynamic Specifications: The Case of PresidentialApproval.” Political Analysis, Vol. 3, 1991: 51–87.

Braun, M. (2001). Why is Fiscal Policy pro-cyclical in Developing Countries? Cam-bridge, MA: Harvard University Press.

Cameron, David R. (1978). “The Expansion of the Public Economy: A ComparativeAnalysis” American Political Science Review, 72, 1243–1261.

Cardoso, Eliana and Ann Helwege (1993). Latin America’s Economy: Diversity,Trends, and Conflicts. Cambridge: MIT Press.

della Paolera, G., M. A. Irigoin and C. Bozzoli (2003). “Passing the buck: Monetaryand Fiscal Policies.” In A. Taylor and G. della Paolera, editors, A new EconomicHistory of Argentina, Cambridge, UK: Cambridge University Press.

Dixit, A., G. M. Grossman and F. Gul (2000). “The Dynamics of Political Compro-

28

mise.” Journal of Political Economy 108(3): 531– 68.

Drazen, A., and V. Grilli (1993). “The Benefit of Crises for Economic Reforms.”American Economic Review 83 (3), pp. 598–607.

Engle, R. and C. Granger (1987). “Cointegration and Error Correction: Representa-tion, Estimation and Testing.” Econometrica Vol. 55: 251–276.

Eslava, M. (2006). The political economy of fiscal policy: survey. Working Paper 583.Washington DC: InterAmerican Development Bank.

Ferreres, O. (2004). Dos siglos de Economıa Argentina (1810– 2004) Historia argentinaen cifras. Buenos Aires: Fundacion Norte y Sur.

Fernandez, R. and D. Rodrik (1991). “Resistance to Reform.” American EconomicReview, Vol. 81 (5): pp. 1146–1155.

Galı, J., and R. Perotti (2003). Fiscal Policy and Monetary Integration in Europe.Economic Policy Vol. 18(37): 533–572.

Gavin, M. and R. Perotti (1997). “Fiscal Policy in Latin America” NBER. Macroeco-nomics Annual 1997.

Gourevitch, Peter A. (1986). Politics in Hard Times: Comparative Responses to In-ternational Economic Crises. Ithaca, NY: Cornell University Press.

Ilzetzki, E. and C. Vegh (2008). Procyclical fiscal policy in developing countries: truthor fiction? NBER Working Paper 14191.

Jaimovich, Dany, and Ugo Panizza (2007). Procyclicality or Reverse Causality? RESWorking Papers 1029. Inter-American Development Bank, Research Department.

Kaminsky, G. L., C. Reinhart and C. Vegh (2004). “When It Rains, It Pours: pro-cyclical Capital Flows and Macroeconomic Policies.” NBER Working Paper 10780.Cambridge, United States: National Bureau of Economic Research.

Kaminsky, G. L., and . M. Reinhart (1999). “The Twin Crises: The Causes of Bankingand Balance-of-Payments Problems.” American Economic Review Vol. 89 (3), pp.473–500.

Katzenstein, P. J., (1985). Small States in World Markets. Ithaca, NY: Cornell Uni-versity Press.

Katzenstein, P J.. (1987) Corporatism and change: Austria, Switzerland, and thepolitics of industry. Ithaca, NY: Cornell University Press.

Kaufman, R. and A. Segura-Ubiergo (2001). “Globalization, Domestic Politics, andSocial Spending in Latin America.” World Politics 53 (July): 553–88.

29

Kontopoulos, Y. and R. Perotti (1999). “Government Fragmentation and Fiscal PolicyOutcomes: Evidence from OECD Countries.” In Fiscal Institutions and Fiscal Per-formance, edited by James Poterba and Jurgen von Hagen, Chicago: University ofChicago Press, pp. 81–102.

Lane, P. (2003) “The cyclical behavior of fiscal policy: evidence from the OECD”Journal of Public Economics

Martinelli, C. and M. Tommasi (1997). “Sequencing of Economic Reforms in thePresence of Political Constraints.” Economics and Politics 9 (2): 115–131.

Mora y Araujo, M. and I. Llorente (1980). El voto peronista. Ensayos de sociologiaelectoral argentina. Buenos Aires: Sudamericana.

Musgrave, R. (1959). The Theory of Public Finance: A Study of Public Economy. NewYork: McGraw-Hill

Nordhaus W. (1975). “The Political Business Cycle.” The Review of Economic Studies,42 (2) pp. 169–190.

North, D. C. (1990). Institutions, Institutional Change and Economic Performance.New York, NY: Cambridge University Press.

Persson, T. and G. Tabellini (2003). The economic effects of constitutions. MIT Press.

Pinto, P. M. (2013). The Politics of Hard Times: Fiscal Policy and the Endogeneityof Economic Recessions. In Miles Kahler and David Lake, editors, Politics in the NewHard Times: The Great Recession and Contemporary Perspective. Ithaca, New York:Cornell University Press

Rodrik, D. (1996). “Understanding Economic Policy Reform.” Journal of EconomicLiterature 34 (1): pp. 9–41.

Rodrik, D. (1998) “Why Do More Open Economies have Bigger Governments.” Journalof Political Economy 106 (5): 997–1032.

Scartascini, C., E. Stein, and M. Tommasi (2010). “Veto Players and Policy Trade-offs.An Intertemporal Approach to Study the Effects of Political Institutions on Policy.”

Sturzenegger, A. and R. Moya (2003). “Economic Cycles.” In G. della Paolera andA. Taylor, editors, A new Economic History of Argentina, Cambridge, UK: CambridgeUniversity Press.

Talvi, E., and C. Vegh (2005) “Tax Base Variability and pro-cyclical Fiscal Policy inDeveloping Countries.” Journal of Development Economics 78(1): 156–190.

Tsebelis G., E. Chang (2004). “Veto players and the structure of budgets in advancedindustrialized countriesd.” European Journal of Political Research, Vol. 43: 449–476.

30

Tornell, A. and P. Lane (1999), “The Voracity Effect” American Economic Review, 89(1); 22–46. March.

Weingast, B., K. Shepsle, and C. Johnsen (1981). “The Political Economy of Benefitsand Costs: A Neoclassical Approach to Distributive Politics.” Journal of PoliticalEconomy 89: 642–64.

Wibbels, E. (2006) Dependency Revisited: International Markets, Business Cycles,and Social Spending in the Developing World. International Organization 60:2 (Spring2006), pp. 433–6

Woo, J. . (2009) “Why Do More Polarized Countries Run More Pro-cyclical FiscalPolicy?” The Review of Economics and Statistics (Forthcoming)

31

Table 1: Government Spending and Output Gap by Decades

Output gap ToT gap Govt. Spend.t−1 Constant N R2

1900-1910 1.486 0.0215** -0.381 5.474 10 0.628(1.725) (0.0078) (0.231) (3.288)

1911-1920 0.944 0.00267 -0.0629 0.862 10 0.205(0.912) (0.0057) (0.127) (1.840)

1921-1930 -0.657 -0.0126* 0.0235 -0.205 10 0.626(1.552) (0.0057) (0.159) (2.346)

1931-1940 1.233 0.00124 -0.180 2.803 10 0.389(1.009) (0.0030) (0.212) (3.261)

1941-1950 3.336* -0.00324 -0.197 3.169 10 0.461(1.688) (0.0042) (0.172) (2.718)

1951-1960 3.117 -0.00493 -0.178 2.929 10 0.403(2.538) (0.0103) (0.455) (7.385)

1961-1970 0.758 0.00990* 0.118 -1.906 10 0.614(0.678) (0.0043) (0.149) (2.477)

1971-1980 -0.779 0.00507 0.117 -1.958 10 0.306(1.669) (0.0034) (0.259) (4.437)

1981-1990 0.161 0.00137 -0.0305 0.484 10 0.019(0.880) (0.0046) (0.279) (4.815)

1991-2004 2.129*** -0.0014 -0.294*** 5.158*** 14 0.840(0.405) (0.0054) (0.073) (1.272)

Standard errors in parentheses; Significance levels: ∗ 1%, ∗∗ 5%, ∗∗∗ 1%

Output gap: natural log deviation of GDP/capita from its Hodrick-Prescott (HP) trend

ToT gap (gap in terms of trade): natural log deviation from HP filtered series

32

Table 2: Error Correction Models: Baseline

Models (1) (2) (3) (4)Constant 0.006 0.008 -0.024 -0.024

(0.011) (0.012) (0.028) (0.028)∆Ln GDP/capita 0.817 ∗∗∗ 0.583 ∗∗ 0.597 ∗∗ 0.576

(0.241) (0.284) (0.275) (0.481)γ (Residualt−1) -0.430 ∗∗∗ -0.414 ∗∗∗ -0.463 ∗∗∗ -0.461 ∗∗∗

(0.090) (0.091) (0.097) (0.098)Peronist -0.013 -0.024 -0.023

(0.022) (0.022) (0.023)Peronist x ∆Ln GDP/cap. 0.895 ∗∗ 0.87 ∗∗ 0.992 ∗∗

(0.446) (0.421) (0.484)ISI 0.061 ∗ 0.058 ∗

(0.035) (0.034)ISI x ∆Ln GDP/cap. 0.145

(0.567)Reform 0.028 0.027

(0.032) (0.032)Reform x ∆Ln GDP/cap. -0.183

(0.600)N 104 104 104 104R2 0.338 0.363 0.395 0.397F 14.62 12.89 10.23 7.65Durbin-Watson d-stat 1.947 1.939 1.934 1.912Significance levels: ∗ 10%, ∗∗ 5%, ∗ ∗ ∗ 1%

DV: ∆Ln Govt. Spend./capita; Huber-White robust std. err. in parenthesis

γ: coeff. on Residualt−1, where Residualst−1 = Gt−1 − θ − δYt−1

ISI: Import Substitution Industrialization; Reform: economic and trade liberalization.

33

Table 3: ECM by administrationVariable Coefficient (Std. Err.)

Constant -0.020 (0.019)∆Ln GDP/cap. 1.434∗∗∗ (0.320)γ (Residualt−1) -0.520∗∗∗ (0.106)Part. Conserv. 0.045 (0.047)Part. Conserv. x ∆Ln GDP/cap. -0.488 (0.735)UCR -0.030 (0.029)UCR x ∆Ln GDP/cap. -0.973∗ (0.562)Military 0.020 (0.028)Military x ∆Ln GDP/cap. -1.203∗ (0.646)Nat. Coal. 0.006 (0.048)Nat. Coal x ∆Ln GDP/cap. -0.398 (2.208)Nat. Dem. 0.059∗∗ (0.024)Nat. Dem x ∆Ln GDP/cap. -0.903∗∗ (0.388)UCRI 0.077 (0.067)UCRI x ∆Ln GDP/cap. -0.218 (0.951)Alianza 0.054∗∗ (0.023)Alianza x ∆Ln GDP/cap. -0.574 (0.544)ISI 0.036 (0.025)N 104R2 0.436F (17,86) 4.446DW d-statistic (18,104) 1.929Significance levels: ∗ 10%, ∗∗ 5%, ∗ ∗ ∗ 1%

DV: ∆Ln Govt. Spend./capita; Huber-White robust std. err. in parenthesis

γ: coefficient on Residualt−1, where Residualst−1 = Gt−1 − θ − δYt−1

Part. Conserv.: 1900-1915; UCR: 1916-1990; 1963-1965; 1983-1989

Nat. Dem: 1932-1937; Nat. Coal: 1938-1942; UCRI: 1958-1962; Alliance: 1999-2001

Military: 1930-1931; 1943-1945 (GOU); 1955-1957; 1966-1972; 1976-1982

Peronist (omitted category): 1946-1955; 1973-1975; 1989-1999; 2002-2004

34

Table 4: ECM models: Peronist, Other Democratic and GOUVariable Coefficient (Std. Err.)

Constant 0.004 (0.015)∆Ln GDP/cap. -0.037 (0.329)γ (Residualt−1) -0.454∗∗∗ (0.093)Peronist -0.008 (0.025)Peronist x ∆Ln GDP/cap. 1.491∗∗∗ (0.478)GOU 0.110∗∗∗ (0.024)GOU x ∆Ln GDP/cap. 1.875∗∗∗ (0.384)Radical -0.017 (0.025)Radical x ∆Ln GDP/cap. 0.544 (0.537)Conservative (All) 0.016 (0.030)Conservative x ∆Ln GDP/cap. 0.891 (0.618)N 104R2 0.414F (10,93) 59.39DW d-statistic(11,104) 1.970DV: ∆Ln Govt. Spend./capita; Huber-White robust std. err. in parenthesis

Significance levels: ∗ 10%, ∗∗ 5%, ∗ ∗ ∗ 1%

Omitted category: Military government

35

Table 5: ECM Models: Alternative hypothesesVariable Coefficient (Std. Err.)

Constant -0.118∗∗ (0.053)∆Ln GDP/cap. -0.130 (0.758)γ (Residualt−1) -0.622∗∗∗ (0.119)Peronist -0.006 (0.029)Peronist x ∆Ln GDP/cap. 1.511∗∗ (0.631)GOU 0.054 (0.047)GOU x ∆Ln GDP/cap. 1.713∗∗ (0.654)Radical 0.053 (0.035)Radical x ∆Ln GDP/cap. 0.827 (0.554)Conservative 0.142∗∗ (0.069)Conservative x ∆Ln GDP/cap. 1.386 (0.945)Nat. Coal. -0.067∗ (0.038)Nat. Coal. x ∆Ln GDP/cap. 2.082 (1.710)Nat. Dem. 0.038 (0.036)Nat. Dem. x ∆Ln GDP/cap. 0.714 (0.614)∆ Ln ToT 0.132 (0.097)∆ Ln US int. rate -0.042∗ (0.024)∆ Ln Imports -0.091 (0.079)ISI (1930-1975) 0.126∗∗ (0.053)ISI x ∆Ln GDP/cap. 0.280 (0.645)Reform (1976-2004) 0.113∗∗ (0.047)Reform x ∆Ln GDP/cap. 0.419 (0.633)N 104R2 0.498F (21,82) 11.087DW d-statistic(22,104) 1.805Significance levels: ∗ 10%, ∗∗ 5%, ∗ ∗ ∗ 1%

DV: ∆Ln Govt. Spend./capita; Huber-White robust std. err. in parenthesis

γ: coefficient on Residualt−1, where Residuals = G− θ −Xδ′

X is the vector of covariates included in the second stage.

Ln US int. rate: Ln US short term interest rate;

Ln ToT: Ln terms of trade; Ln Imports: Ln import quantities

Omitted category: Military government

36

Table 6: ECM Models: decade dummiesVariable Coefficient (Std. Err.)

Constant 0.002 (0.037)∆Ln GDP/cap. 0.225 (0.719)γ (Residualt−1) -0.646∗∗∗ (0.133)Peronist -0.016 (0.035)Peronist x ∆Ln GDP/cap. 1.582∗∗ (0.678)GOU 0.055 (0.087)GOU x ∆Ln GDP/cap. 1.621∗∗ (0.808)Radical 0.044 (0.034)Radical x ∆Ln GDP/cap. 0.567 (0.701)Conservative 0.176∗∗∗ (0.063)Conservative x ∆Ln GDP/cap. 0.702 (0.946)Nat. Coal. -0.074 (0.068)Nat. Coal. x ∆Ln GDP/cap. 2.052 (1.728)National Dem. 0.024 (0.068)National Dem. x ∆Ln GDP/cap. 0.617 (0.780)∆ Ln ToT 0.111 (0.100)∆ Ln US int. rate -0.040 (0.031)∆ Ln Imports -0.089 (0.084)N 104R2 0.557F (26,77) 10.682DW d-statistic(27,104) 1.942DV: ∆Ln Govt. Spend./capita; Huber-White robust std. err. in parenthesis

Includes decade dummies; excluded cat.: 1990-2004

See table 3 for references

Significance levels: ∗ 10%, ∗∗ 5%, ∗ ∗ ∗ 1%

Omitted category: Military government

37

Table 7: Instrumental Variable Regression (2SLS)

(1) (2) (3)Output Gap 0.005 0.004 0.004

(0.004) (0.005) (0.005)Output Gap x Peronist 0.014** 0.016** 0.016**

(0.006) (0.007) (0.007)Peronist -0.012 -0.010 -0.016

(0.020) (0.021) (0.022)US interest rate 0.003 0.001

(0.003) (0.003)∆ToT 0.0003 0.0003

(0.0009) (0.0009)ISI (1930-1975) 0.001

(0.031)Reform (1976-2004) 0.014

(0.032)Constant 0.004 -0.007 -0.006

(0.012) (0.019) (0.031)N 105 104 104R2 0.268 0.259 0.262F -statistic (first stage) 16.64 9.84 9.41DW d-statistic (p-value) 0.358 0.323 0.339Standard errors in parentheses

Significance levels: ∗ 1%, ∗∗ 5%, ∗∗∗ 1%

DV: Govt. spend./capita deviation from Hodrick-Prescott (HP) trend

Output gap: GDP/capita deviation from Hodrick-Prescott (HP) trend

38

Appendix ITable A: Argentina: Government orientation

Presidents Tenure Party Regime Ideology Support base

J.A. Roca 1898-1904 PAN Dem open/agro-exp Agr.producersM.Quintana 1904-1906 PAN Dem open/agro-exp Agr.producersJ.Figueroa Alcorta 1906-1910 PAN Dem open/agro-exp Agr.producersR.Saenz Pena 1910-1914 Union Nacional Dem open/agro-exp Agr.producersV. De la Plaza 1914-1916 PC Dem open/agro-exp Agr.producersH. Yrigoyen 1916-1922 UCR Dem∗ open/agro-exp Middle classesM.T. de Alvear 1922-1928 UCR Dem open/agro-exp Middle classesH. Yrigoyen 1928-1930 UCR Dem open/agro-exp Middle classesJ.F. Uriburu 1930-1932 MILITARY Aut open/agro-exp Agr.producers

A.P. Justo 1932-1938 National Dem Dem† open/agro-exp Agr.producers/industrialistsR.M Ortiz 1938-1942 National Coaltion Dem open/agro-exp Agr.producers/industrialistsR.S Castillo 1942-1943 National Coaltion Dem open/agro-exp Agr.producers/industrialistsA.Rawson 1943 MILITARY Aut closed/ISI Urban workers-industrialistsP. Ramirez 1943-1944 MILITARY Aut closed/ISI Urban workers-industrialistsE.J. Farrell 1944-1946 MILITARY Aut closed/ISI Urban workers-industrialistsJ.D. Peron 1946-1955 PJ Dem∗∗ closed/ISI Urban workers-industrialistsE. Lonardi 1955 MILITARY Aut closed/ISI Agr. ProducersP. Aramburu 1955-1958 MILITARY Aut closed/ISI Agr. ProducersA. Frondizi 1958-1962 UCRI Dem‡ closed/ISI Middle classesJ.M. Guido 1962-1963 UCRI Aut closed/ISI Middle classesA. Illia 1963-1966 UCRP Dem‡ closed/ISI Middle classesJ.C. Ongania 1966-1970 MILITARY Aut closed/ISI Agr.producersR. Levingston 1970-1971 MILITARY Aut closed/ISI Agr.producersA. Lanuse 1971-1973 MILITARY Aut closed/ISI Agr.producersH. Campora 1973 PJ Dem closed/ISI Urban workers-industrialistsR.Lastiri 1973 PJ Dem closed/ISI Urban workers-industrialistsJ.D. Peron 1973-1974 PJ Dem closed/ISI Urban workers-industrialistsI. de Peron 1974-1976 PJ Dem closed/ISI Urban workers-industrialistsJ. Videla 1976-1981 MILITARY Aut open/reform Agr.producers/contractorsR. Viola 1981 MILITARY Aut open/reform Agr.producers/contractorsL.Galtieri 1981-1982 MILITARY Aut open/reform Agr.producers/contractorsR. Bignone 1982-1983 MILITARY Aut open/reform Agr.producers/contractorsR. Alfonsin 1983-1988 UCR Dem partial open Middle classesC. Menem 1989-1999 PJ Dem open/reform Urban workers-industrialistsA.De la Rua 1999-2001 UCR-Alliance Dem open/reform Middle classesE.Duhalde 2002 PJ Dem open/reform Urban workers-industrialistsN.Kirchner 2003-2007 PJ Dem partial open Urban workers-industrialistsC. Kirchner 2007– PJ Dem partial open Urban workers-industrialists

Source Mollinelli et al 1999; authorsReferences: Military: follows a coup; PAN/PC=Conservative Party PJ=Peronists UCR=RadicalsUCRI=UCR Intransigente UCRP= UCR del Pueblo† The UCR was proscribed and banned from the 1931 election; pervasive electoral fraud through 1942.∗ First presidential election under universal, mandatory and secret ballot (Ley Saenz Pena of 1912∗∗ Female suffrage instituted in 1947. The first Presidential election with female suffrage was 1952.‡ The Peronist party was proscribed and banned from the 1958 and 1963 elections

Table B: Argentina: Economic Recessions 1903-2004

Highest Economic Length Drop in GDP Drop in industry Drop in construction Annual inflationyear cycle (years) (start to end) (start to end) (start to end)

1913 1913-1917 4 -0.200 -0.167 -0.824 0.0791929 1929-1932 3 -0.151 -0.176 -0.594 -0.0791951 1951-1952 1 -0.060 -0.093 -0.101 0.3861958 1958-1959 1 -0.061 -0.129 -0.159 1.1371961 1961-1963 2 -0.037 -0.094 -0.156 0.5891974 1974-1976 2 -0.011 -0.055 0.202 14.3851977 1977-1978 1 -0.036 -0.105 -0.048 1.7551980 1980-1982 2 -0.075 -0.143 -0.207 1.3271984 1984-1985 1 -0.069 -0.149 -0.149 6.7221987 1987-1990 3 -0.108 -0.150 -0.498 14.0371994 1994-1995 1 -0.036 -0.072 -0.122 0.0341998 1998-2002 4 -0.184 -0.270 -0.508 0.219

Highest Economic Length Initial shock Initial shock Drop in imports Real devaluationyear cycle (years) Export prices Export volumes (start to end) (implicit)

1913 1913-1917 4 0.011 -0.262 -0.5581929 1929-1932 3 -0.110 -0.317 -0.527 0.6611952 1951-1952 1 -0.166 -0.295 -0.403 0.1701959 1958-1959 1 0.046 -0.038 -0.153 0.2431961 1961-1963 2 -0.169 -0.082 -0.322 0.1061974 1974-1976 2 0.249 -0.166 -0.175 1.2001977 1977-1978 1 0.064 0.096 -0.129 -0.2551980 1980-1982 2 0.180 0.110 -0.525 3.2291984 1984-1985 1 0.086 0.208 -0.171 0.1601987 1987-1990 3 -0.177 -0.099 -0.378 0.7501994 1994-1995 1 0.029 0.249 -0.116 0.0001998 1998-2002 4 -0.104 -0.008 -0.676 1.400

Source: CEPAL Buenos Aires, FIEL and authors